Nkpado egwu na Transformers
Nkpado egwu na-eji ụdị mgbanwe iji gee egwu na ịkọ nkọwa nkọwa dị ka ụdị, ọnọdụ, ngwa na oge.
Nchịkọta
Ọ na-akwado ọchụchọ, ntụnye, na nhazi akpaaka n'ofe nnukwu katalọgụ egwu.
Ime miri emi
Nkpado egwu akpaaka bụ nsogbu nhazi akara ọtụtụ: otu egwu nwere ike ịbụ 'oke nkume,'' ike,' 'guitar' na 'ngwaọrụ' otu mgbe. Ndị na-eme mgbanwe na-eme ya site n'ịtụgharị ọdịyo ka ọ bụrụ spectrogram (onyinyo ugboro ugboro) na inye nri ya site na nlebara anya onwe ya, dị ka ihe ntụgharị ọhụụ na-emeso patches oyiyi. Ụdị dị ka Audio Spectrogram Transformer (AST) na MERT na-amụta ụkpụrụ ogologo n'ofe egwu dum, na-ewere ka otu egwu egwu si metụta nkeji nkeji dịpụrụ adịpụ. Ọtụtụ ndị a zụrụ nke ọma na-elekọta onwe ha na nde vidiyo na-enweghị aha, wee na-ekiri ya nke ọma na ntanetị mkpado dị ka MagnaTagaTune ma ọ bụ Ndekọ Song Nde. N'ihi na mkpado anaghị ekewapụ onwe ya, oyi akwa ikpeazụ na-eji nsonaazụ sigmoid gbagoro megide akara ngosi dị ka nkenke nkenke yana ROC-AUC.
Nghọta nka nka
A na-atụgharị ụda olu raw ka ọ bụrụ log-Mel spectrogram, kewaa n'ime patches na-ekpuchi ma tinye ya n'ahịrị na ntinye ọnọdụ. Nlebara anya onwe onye na-eme ka patch ọ bụla tụọ ihe nkedo ọ bụla, yabụ mmemme egwu dị anya na-emetụta mkpado ọ bụla. N'adịghị ka nhazi ọkwa onyonyo nke otu akara, mkpado egwu na-emetụta sigmoid kwa mkpado karịa otu softmax, ebe ọ bụ na aha na-emekọ ọnụ. Ọzụzụ izizi nke onye na-elekọta onwe ya (ịkọ amụma ụda ọdịyo kpuchiri ekpuchi) na-enye ihe ngosi siri ike tupu emezigharị ya na obere ihe akpọrọ.
Mmetụta atụmatụ
Nweta na iru
Ọ na-eme ka nnweta ya dịkwuo mma site na ndegharị, ịkọ akụkọ, na ntụgharị olu.
Ọnụ ego na mmefu ego
Ndị otu mgbasa ozi nwere ike ibubata ọdịyo a na-egbu maramara ngwa ngwa site na iji obere mmefu ego.
Ọsọ na ọnụ ọgụgụ
Sistemụ na-eche ihu ndị ahịa nwere ike hazie mkparịta ụka n'ọtụtụ buru ibu.
Ọdịnihu nke egwu mkpado na Transformers
Mkpado na-ejikọta na nghọta asụsụ okike ka ị nwee ike chọọ 'nrọ lo-fi na vinyl crackle maka ọmụmụ' kama bọtịnụ ụdị edozi. Ụdị ederede ọdịyo dị iche iche dị ka CLAP kwadoro egwu na nkọwa n'otu oghere, na-eme ka mkpado efu na-ahụtụbeghị na ọzụzụ. Na-atụ anya ka ịba ụba, akara granular karịa, njikwa ụdị ngwakọta ka mma, yana mkpado ngwaọrụ maka nzuzo. Mkparịta ụka ikike na njiri mara gbasara ọzụzụ na katalọgụ nwebiisinka ga-akpụzi data ụdịdị ndị a nwere ike iji.
Mmejuputa n'ezie n'ụwa
Ụdị na-emepụta akpaaka na mkpado ọnọdụ ka ọrụ mgbasa ozi nwee ike wulite listi ọkpụkpọ 'focus' ma ọ bụ 'mmega ahụ'
Ikwe ka ọba akwụkwọ egwu na-ebuli egwu egwu 'upbeat acoustic guitar' maka ndị ndezi vidiyo na-achọ ikikere mmekọrịta.
Igwe nkwado na-enye ike na-ahụ egwu ndị yiri ya na-enweghị atụ karịa ihe ndị ọrụ gosipụtara n'ụzọ doro anya
Ịhazi mkpokọta ihe atụ nke onye nrụpụta site na ngwa, igodo, na tempo achọpụtara na-akpaghị aka
Ihe ize ndụ & okporo ụzọ nche
Iji olu eme ihe na ihe egwu mpụta ga-abawanye mgbe nkwenye na-efu.
Izi ezi nwere ike ịdaba n'ofe ụda olu, olumba ma ọ bụ gburugburu mkpọtụ.
Enwere ike imehie ọdịyo sịntetik dị ka ezigbo okwu na-enweghị akara doro anya.
Map mmejuputa
Nweta nkwenye doro anya maka ijide olu, imechi, na ijigharị.
Nwale ogo n'ofe ndị na-ekwu okwu dị iche iche yana ọnọdụ ndabere.
Kọwaa mgbe mmadụ ga-enyocha ma ọ bụ kwado nsonye.
Deba aha ọdịyo sịntetik ma debe ndekọ ihe ndekọ maka ịza ajụjụ.
Nọgide na-eme nchọpụta
Free newsletter
Get the daily AI briefing
Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.
One email each weekday. Unsubscribe in one click. We never sell or share your address.
Test yourself
Take the Music Tagging with Transformers quiz
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
Ntuziaka na-esote
Egwu na-akpado akpaaka
Ajụjụ a na-ajụkarị
Kedu ihe bụ mkpado egwu na Transformers?
Nkpado egwu na-eji ụdị mgbanwe iji gee egwu na ịkọ nkọwa nkọwa dị ka ụdị, ọnọdụ, ngwa na oge. Ọ na-akwado ọchụchọ, nkwanye, na nhazi akpaaka n'ofe nnukwu katalọgụ egwu.
Kedu ihe kpatara eji ewere mkpado egwu dị ka nsogbu ọtụtụ akara?
Egwu nwere ike ịbụ okwute, ike na ụbọ akwara n'otu oge, yabụ ọtụtụ mkpado na-emetụta otu egwu.
Kedu ihe nnochite anya ihe nrịbama transformer na-ejikarị?
A na-atụgharị ụda ka ọ bụrụ spectrogram ugboro ugboro, wee mechie ma nye ya nri site na nlebara anya onwe onye dị ka patches oyiyi.
Kedu ihe kpatara ụdị mkpado egwu ji eji mmepụta sigmoid kwa mkpado kama otu softmax?
Softmax na-amanye puru omume iji chịkọta otu (otu nhọrọ); sigmoid na-ahapụ mkpado ọ bụla bụrụ eziokwu n'onwe ya.
Kedu ọrụ nke ọzụzụ ọzụzụ nke onwe onye maka ụdị dịka MERT?
Ịma ụzọ na amụma ọdịyo kpuchiri ekpuchi n'elu nnukwu ụlọ ọrụ na-enweghị akara na-ewulite ihe nnochite anya ma emesịa nke ọma na-ekiri na data akara.
Kedu ihe ndekọ data a na-ejikarị eme ka ọ dị mma na mkpado egwu benchmark?
MagnaTagaTune na Ndekọta Song Dataset bụ akara akara egwu egwu; MNIST na ImageNet bụ nhazi onyonyo.